Serpentine minichannel liquid-cooled heat sinks for electronics cooling applications

(1)

Serpentine Minichannel Liquid-Cooled Heat Sinks for

Electronics Cooling Applications

By

Ahmed Fouad Mahmood Al-Neama

Submitted in accordance with the requirements for the degree of

Doctor of Philosophy

The University of Leeds

School of Mechanical Engineering

Institute of ThermoFluids (iTF)

January 2018

The candidate confirms that the work submitted is his own, except where work formed jointly-authored publication has been included. The contribution of the candidate and

the other authors to this work has been explicitly indicated overleaf. The candidate confirms that appropriate credit has been given within the thesis where reference has

been made to the work of others.

This copy has been supplied on the understanding that is copyrighted material and that no quotation from the thesis may be published without proper acknowledgement.

(2)

[i]

Work Formed from Jointly Authored Publications

The candidate has publications from work contained in this thesis, and these were submitted with the support of supervisors. The contribution of the candidate to these works is explicitly mentioned below.

1- The work in chapters 4, 5 and 6 of the thesis has appeared in publication as follows: Applied Thermal Engineering, 2017, Ahmed F. Al-Neama, Nik Kapur, Jonathan L. Summers, Harvey M. Thompson. The title of the paper is “An experimental and numerical investigation of the use of liquid flow in serpentine microchannels for microelectronics cooling”. 116, pp. 709–723. (Published

online)

2- The work in part of chapters 4 and 6 and chapter 7 of the thesis has appeared in publication as follows: International Journal of Heat and Mass Transfer, 2018, Ahmed F. Al-Neama, Zinedine Khatir, Nikil Kapur, Jonathan Summers, Harvey M. Thompson. The title of the paper is “An experimental and numerical investigation of chevron fin structures in serpentine minichannel heat sinks”. 120, pp. 1213–1228. (Published online)

3- The work in part of chapters 4, 6 and 7 and chapter 8 of the thesis has appeared in publication as follows: Applied Thermal Engineering, 2018, Ahmed F. Al-Neama, Nik Kapur, Jonathan L. Summers, Harvey M. Thompson. The title of the paper is “Thermal management of GaN HEMT devices using serpentine minichannel heat sinks”. (In press)

The candidate has conducted the majority of the work that appears in the published articles, such as developing the models, collecting the experiment data, optimisation, and presenting and analysing the results. The co-authors as appears in the published papers provided valuable review, contribution and guidance to the candidate.

(3)

[ii]

Acknowledgements

I would like to express my deepest gratitude to my supervisors Prof. Nikil Kapur, Prof. Harvey M. Thompson and Dr. Jonathan L. Summers for their advice, assistance, supportive guidance, and encouragement throughout the entire research.

I would also like to acknowledge the Iraqi Ministry of Higher Education and Scientific Research (MOHE), and Mechanical Engineering Department University of Mosul, Iraq, who offered unlimited help and financial support.

I am most thankful to my mother and father, wife, children, brothers and all my family members for their help, support, encouragement and patience during the years of my study.

Finally, I am grateful to University of Leeds for granting me access to their facilities, such as lab, library and Advanced Research Computing (ARC2).

Above all, I would like to thank the Almighty God for granting me the power and patient to accomplish this work. Without the willing of Allah, nothing is possible.

(4)

[iii]

Abstract

The increasing density of transistors in electronic components is leading to an inexorable rise in the heat dissipation that must be achieved in order to preserve reliability and performance. Hence, improving the thermal management of electronic devices is a crucial goal for future generations of electronic systems. Therefore, a complementary experimental and numerical investigation of single-phase water flow and heat transfer characteristics of the benefits of employing three different configurations of serpentine minichannel heat sink (MCHS) designs has been performed, to assess their suitability for the thermal management of electronic devices. These heat sinks are termed single (SPSMs), double (DPSMs) and triple path serpentine rectangular minichannels (TPSMs), and their performance is compared, both experimentally and numerically, with that of a design based on an array of straight rectangular minichannels (SRMs) in terms of pressure drop (∆𝑃), average Nusselt number (𝑁𝑢𝑎𝑣𝑔) and total thermal resistance (𝑅𝑡ℎ).

The results showed that the serpentine channel bends are very influential in improving heat transfer by preventing both the hydrodynamic and thermal boundary layers from attaining a fully-developed state. The SPSM design provides the most effective heat transfer, followed by the DPSM and TPSM ones, both of which out-performed the SRM heat sink. The SPSM heat sink produced a 35% enhancement in 𝑁𝑢𝑎𝑣𝑔 and a 19% reduction in 𝑅𝑡ℎ at a volumetric flow rate

(𝑄𝑖𝑛) of 0.5 𝑙/𝑚𝑖𝑛 compared to the conventional SRM heat sink. These improvements in the heat

transfer are, however, achieved at the expense of significantly larger ∆𝑃.

It was found that the incorporation of serpentine minichannels into heat sinks will significantly increase the heat-removal ability, but this must be balanced with the pressure drop requirement. Therefore, an experimental and numerical investigation of the benefit of introducing chevron fins has been carried out to examine the potential of decreasing pressure drop along with further thermal enhancement. This novel design is found to significantly reduce both the ∆𝑃 across the heat sink and the 𝑅𝑡ℎ by up to 60% and 10%, respectively, and to enhance the 𝑁𝑢𝑎𝑣𝑔 by 15%,

compared with the SPSM heat sink without chevron fins.

Consequently, the design of the SPSM with and without chevron fins was then optimised in terms of the minichannel width (𝑊𝑐ℎ), number of minichannels (𝑁𝑐ℎ) and chevron oblique angle (𝜃).

The optimisation process uses a 30 (without chevron fins) and 50 (with chevron fins) point Optimal Latin Hypercubes Design of Experiment, generated from a permutation genetic algorithm, and accurate metamodels built using a Moving Least Square (MLS) method. A Pareto front is then constructed to enable the compromises available between designs with a low pressure drop and those with low thermal resistance to be explored and appropriate design parameters to be chosen. These techniques have then been used to explore the feasibility of using serpentine MCHS and heat spreaders to cool GaN HEMTs.

(5)

[iv]

List of Tables

Table 1.1: Comparison between GaN and other Semiconductor properties... 4

Table 1.2: Channel classification by Mehendale et al. (2000) and Kandlikar and Grande (2003)... 7

Table 2.1: Boiling regimes. ... 61

Table 3.1: Flow types. ... 76

Table 3.2: Values of Incrementał pressure drop, Poiseuille number and constant C for rectangular channel at different aspect ratio (Shah, 1978). ... 82

Table 3.3: Fully developed laminar flow Nusselt number in rectangular channel for the H1 boundary condition with three and four walls transferring heat, and with uniform axial heat flux (Phillips, 1987). ... 85

Table 3.4: Developing laminar flow Nusselt number in rectangular channel for the H1 boundary condition used in the computer simulation of Phillips (1987 & 1990)... 85

Table 4.1: Specification of parameters of four MCHSs. ... 101

Table 4.2: Surface roughnesses of four MCHSs. ... 108

Table 4.3: Measurement uncertainty summary. ... 126

Table 5.1: Calculated thermal entrance length for the straight rectangular MCHS test section at different Reynolds number and input power of 100 W. ... 136

Table 6.1: The boundary conditions of the conjugate heat transfer model. ... 157

Table 6.2: Grid dependency test. ... 160

Table 6.3: Summary of mesh properties and calculated values used in mesh independence tests. ... 162

Table 6.4: Minichannel design performance at six operating conditions points located on the Pareto front together with CFD validation as shown in Fig. 6.37. ... 181

Table 7.1: Dimensional details for SMPF and SMCF heat sinks. ... 184

Table 7.2: Uncertainty for various critical parameter of serpentine MCHSs. ... 185

Table 7.3: Grid dependency tests. ... 191

Table 7.4: Details of MCHS dimensions with oblique fin used for validation, all dimensions in mm. ... 192

Table 7.5: Minichannel design performance at six operating conditions points located on the Pareto front together with CFD validation as shown in Fig. 7.29. ... 211

Table 8.1: Thickness and thermal conductivity of the materials used for simulation. ... 217

Table 9.1: Minichannel design performance at six operating conditions points located on the Pareto front. ... 230

(10)

[ix]

List of Figures

Fig. 1.1: shows the increase in the number of transistors every year. The blue line represents Moore’s law applied, starting from the first microprocessor on the graph. The red line denotes the least square fit of the entire set (Saenen, 2013). ... 1 Fig. 1.2: Typical convective thermal resistances for a 10 cm2 heat source area (Tummala, 2001). ... 3 Fig. 1.3: Schematic illustration of a microchannel heat sink with enlarged microchannel. ... 6 Fig. 1.4: Single microchannel with thermal resistance networks. ... 9 Fig. 1.5: 3-D schematic view of MCHS suggested; a) SRM; b) SPSM; c) DPSM, d) TPSM heat sink. ... 10 Fig. 1.6: (a) CGHV1J070D GaN-on-SiC HEMT bare die which is inside (b); (b) High voltage GaN HEMT on SiC substrate packaging which is inside (c); (c) Radio Frequency (RF) Power Amplifier (Cree Inc., 2012). ... 11 Fig. 1.7: 3-D schematic view of MCHS suggested; a) single path serpentine MCHS with plate fins; b) single path serpentine MCHS with chevron fins. ... 12 Fig. 1.8: Flow chart to explain the scope of the present thesis work. ... 13 Fig. 2.1: 3-D view of a stacked MCHS design (Wei and Joshi, 2003). ... 18 Fig. 2.2: Various inlet/outlet flow arrangements for: (a) I-type; (b) Z-type; (c) ]-type; (d) L-type; and (e) Γ-type (Lu and Wang, 2006). ... 29 Fig. 2.3: Typical inlet/outlet configurations for S, D, N, U and V-arrangement of a straight rectangular microchannel heat sink (Chein and Chen, 2009). ... 30 Fig. 2.4: Different flow arrangements in a straight rectangular microchannel heat sink: (a) U-type; (b) S-type; and (c) P-type (Sehgal et al., 2011). ... 31 Fig. 2.5: Schematic diagram of heat sink configurations (Vinodhan and Rajan, 2014). ... 31 Fig. 2.6: 3-D schematic view of micro pin fin heat sink suggested by Peles et al. (2005). .. 35 Fig. 2.7: 3-D rendering of assembled microchannel cooler (Colgan et al., 2007). ... 36 Fig. 2.8: Effect of nanofluid volume fraction on Nusselt number at different Reynolds number (Seyf and Feizbakhshi, 2012). ... 38 Fig. 2.9: Online and offset micro pin-fin heat sinks (Rubio-Jimenez et al., 2013). ... 39 Fig. 2.10: Dimensions of the test device with enlarge for the PPF arrays and silicon-based microchannel with PPFs (half of the channel) (Yu et al., 2016a). ... 40 Fig. 2.11: Cross section view of wavy MCHS (Sui et al., 2010)... 43 Fig. 2.12: Velocity vectors for serpentine wavy microchannel; a) A=50 µm; and b) A=200 µm (Gong et al., 2011). ... 44 Fig. 2.13: Velocity vectors for raccoon wavy microchannel; a) A=50 µm; and b) A=150 µm (Gong et al., 2011). ... 44 Fig. 2.14: A schematic view of a heat sink unite with (a) zigzag channels; (b) curvy

channels; and (c) step channels (Mohammed et al., 2011c). ... 45 Fig. 2.15: A schematic view of a swirl MCHS with (a) straight channels; (b) Cω3; and (c) Cω5 (Fan and Hassan, 2011). ... 45

(11)

[x]

flow passages (Yong and Teo, 2014). ... 47 Fig. 2.17: Numerical results for streamline plots at 𝑅𝑒 =100 and 𝑅𝑒 =200 (Yong and Teo, 2014). ... 47 Fig. 2.18: Local Nusselt numbers of a wavy and a wavy microchannel with dimples (Gong et al., 2016). ... 48 Fig. 2.19: Schematic view for MCHS proposed by Steinke and Kandlikar (2004) for: (a) Secondary flow channels; and (b) Venturi based secondary flow. ... 49 Fig. 2.20: Cross section view for a MCHS used by Xu et al. (2005). ... 50 Fig. 2.21: Cross section view of oblique fins suggested by Lee et al. (2012 and 2013). ... 51 Fig. 2.22: Straight rectangular, fan-shaped and triangular reentrant cavities microchannels (Chai et al., 2013a), all dimensions in mm. ... 52 Fig. 2.23: Thermal resistance versus pumping power (Chai et al., 2013a). ... 52 Fig. 2.24: Schematic diagram of computational domain for reentrant microchannels (Deng et al., 2015a). ... 53 Fig. 2.25: Schematic diagrams of (a) internal Y-shaped bifurcation microchannel; (b) vertical fin without Y-shaped bifurcation microchannel; (c) The cross profiles of

microchannel heat sinks with Y-shaped bifurcation at angle of 120o_{and 180}o_{(Xie et al.,}

2015). ... 55 Fig. 2.26: Structure and dimensions of offset ribs suggested by Chai et al. (2016a) (all dimensions in mm). ... 57 Fig. 2.27: Structure and dimensions of offset ribs proposed by Chai et al. (2016b) (all dimensions in mm). ... 58 Fig. 2.28: (a) Schematic diagrams of SCRR; and (b) geometric parameters of MC-SCRR (Ghani et al., 2017b). ... 58 Fig. 2.29: (a) Schematic diagrams of SOCRR; and (b) geometric parameters of MC-SOCRR (Ghani et al., 2017c). ... 59 Fig. 2.30: Typical features of a boiling curve for water at 1 atmospheric pressure (Bergman et al., 2017). ... 61 Fig. 2.31: (a) Schematic illustration of the heat sink device used by Koşar et al. (2006); (b) flow distributive pillars; (c) geometry of the inlet orifices (dimensions in µm). ... 64 Fig. 2.32: Different inlet/outlet configurations investigated by Wang et al. (2008). ... 65 Fig. 2.33: Schematic illustration for different microchannel configurations proposed by Li et al. (2017); (a) conventional microchannel (R); and (b) microchannel with triangular cavities (TC). ... 67 Fig. 3.1: Simple flowchart for types of convective heat transfer. ... 72 Fig. 3.2: Prandtl number versus temperature for various fluids (Nellis and Klein, 2009). .. 74 Fig. 3.3: The hydrodynamic and thermal boundary layer thicknesses for (a) 𝑃𝑟 > 1; and (b) 𝑃𝑟 < 1. ... 74 Fig. 3.4: Developing flow (hydrodynamically and thermally developing flow), followed by fully developed (hydrodynamically and thermally fully developed flow) in channel under constant wall temperature boundary condition (Shah and Bhatti, 1987). ... 75

(12)

[xi]

Fig. 3.5: Velocity boundary layer development on a flat plate (Bergman et al., 2017). ... 77

Fig. 3.6: Variation of velocity boundary layer thickness and the local heat transfer coefficient for flow over an isothermal flat plate (Bergman et al., 2017). ... 78

Fig. 3.7: Control volume of fluid flowing in a tube... 78

Fig. 3.8: Axial temperature variation for heat transfer in a channel; (a) Constant surface heat flux; and (b) Constant surface temperature. ... 79

Fig. 3.9: Developing laminar flow apparent friction factor in rectangular channel employed in computer simulation of Phillips (1987). ... 83

Fig. 3.10: Schematic diagram showing the thermal resistance components (Phillips, 1990). ... 90

Fig. 3.11: Description of definition used for calculating error measures. ... 97

Fig. 4.1: Schematic diagram of the experiment setup. ... 99

Fig. 4.2: Photograph of the experimental facility. ... 100

Fig. 4.3: Model and actual pictures of the straight rectangular MCHS test section, all dimensions in mm. ... 101

Fig. 4.4: Model and actual pictures of the (a) single, (b) double and (c) triple path multi-serpentine rectangular MCHS test section respectively from top to bottom, all dimensions in mm. ... 102

Fig. 4.5: (a) SolidWorks model; and (b) actual picture for the single path serpentine rectangular MCHS assembly. ... 103

Fig. 4.6: (a) SolidWorks model; and (b) actual picture for the multi-straight rectangular MCHS assembly. ... 104

Fig. 4.7: Cross-sectional view of the SRM design to explain the thermocouples location. All dimensions in mm. ... 105

Fig. 4.8: Photograph of the power film resistor, all dimensions in millimeter. ... 106

Fig. 4.9: Schematic diagram to describe the junction temperature measurement technique. ... 106

Fig. 4.10: Photograph of (a) applying thermal Ethoxy adhesive to the loading area of the device; and (b) After applied resistors on the Ethoxy adhesive. ... 107

Fig. 4.11: Photographs of surface roughness measurements for (a) SRM; (b) SPSM; (c) DPSM; and (d) TPSM heat sinks. ... 109

Fig. 4.12: Photographs of type-K thermocouples. ... 110

Fig. 4.13: (a) Schematic showing the placement of inlet thermocouple; and (b) Picture of the connection. ... 111

Fig. 4.14: A typical thermocouple calibration curve. ... 112

Fig. 4.15: Rotameter calibration curve for the inlet water to the MCHS. ... 113

Fig. 4.16: Pressure gauge calibration curve results. ... 113

Fig. 4.17: The time history of experimental temperature measured at each thermocouple for the SRM heat sink at total input power of 100 W and volumetric flow rate of 0.10 l/min. 115 Fig. 4.18: Schematic representation for SRM design showing the thermocouple locations and pressure drop components. All dimensions in mm. ... 118

(13)

[xii]

manifolds and the microchannels: (A) ∝ < 0.1; and (B) 0.1 ≤∝≤ 1.0 (Kays and London, 1984). ... 119 Fig. 4.20: Schematic representation of the geometry for the curved rectangular minichannel. ... 123 Fig. 4.21: Schematic representation of the geometry for the serpentine rectangular

minichannel. ... 124 Fig. 4.22: Cross-section view of the test section to explain the arrangement of the wall temperature measurement. ... 125 Fig. 5.1: The total pressure drop in a straight rectangular MCHS model versus 𝑄_𝑖𝑛 at three different input powers of 50, 75 and 100 W. ... 127 Fig. 5.2: Density and kinematic viscosity of water at different temperature. ... 128 Fig. 5.3: Comparison of the total experimental pressure drop with Eq. (4.17) for SRM heat sink model at different 𝑄_𝑖𝑛 for input power of 100 W. ... 128 Fig. 5.4: The experimental minichannel pressure drop versus 𝑅𝑒 at three different input powers of 50 W, 75 W and 100 W. ... 129 Fig. 5.5: The minichannel friction factor versus 𝑅𝑒 at input power of 100 W. ... 130 Fig. 5.6: The experimental Poiseuille number against dimensionless length for rectangular MCHS test section at input power of 100 W. ... 131 Fig. 5.7: The total experimental pressure drop for: a) SPSM heat sink; b) DPSM & TPSM heat sinks at different 𝑄_𝑖𝑛 and input power of 100 W. ... 132 Fig. 5.8: Comparison of experimental data and energy balance predictions for water

temperature rise from MCHS inlet to outlet at two input power of 50 and 100 W. ... 133 Fig. 5.9: Thermocouple temperature readings at different locations inside MCHS versus Reynolds number for input power of a) 50 W and b) 100 W. ... 134 Fig. 5.10: The experimental local and average heat transfer coefficient at different locations versus Reynolds number for input power of 100 W. ... 135 Fig. 5.11: The local average heat transfer coefficient versus the non-dimensional thermal length for input power of 100 W. ... 137 Fig. 5.12: The average Nusselt number versus 𝑅𝑒 for input power of 100 W. ... 138 Fig. 5.13: The average Nusselt number versus 𝑄_𝑖𝑛 for four MCHS models at input power of 100 W. ... 139 Fig. 5.14: Thermal resistance versus volumetric flow rate at 100 W input power. ... 140 Fig. 5.15: Performance evaluation criterion obtained from experiments for three serpentine MCHS designs versus 𝑄_𝑖𝑛 at input power of 100 W. ... 141 Fig. 5.16: Repeatability of 𝑁𝑢𝑎𝑣𝑔 and ∆𝑃𝑡 measurements at different volumetric flow rate and total input power of 100 W. ... 142 Fig. 6.1: Typical finite element mesh. ... 145 Fig. 6.2: Main stages in a CFD simulation. ... 145 Fig. 6.3: Types of elements. The source

(14)

[xiii]

Figure). ... 147

Fig. 6.5: Distance 𝛿𝑤 from wall to computational domain. ... 148

Fig. 6.6: Mesh quality for the TPSM heat sink: a) Laminar flow; b) Turbulent flow. ... 149

Fig. 6.7: Mesh quality for the SRM heat sink: a) Laminar flow; b) Turbulent flow. ... 150

Fig. 6.8: Schematic diagram for conjugate heat transfer between solid and fluid domains. ... 155

Fig. 6.9: 3D geometry with boundary conditions of (a) the straight rectangular MCHS; (b) single path serpentine MCHS. ... 158

Fig. 6.10: Numerical mesh using grid 3 for a TPSM design. ... 160

Fig. 6.11: Schematic diagram of the straight rectangular microchannel with symmetry planes and boundary conditions. (all dimensions in µm) ... 161

Fig. 6.12: Local grid structure on the straight rectangular microchannel for: (a) hexahedral (structured); and (b) tetrahedral (unstructured) meshes... 162

Fig. 6.13: Velocity vector (m/s) distribution along the streamwise on the 𝑥 − 𝑦 section at 𝐻_𝑐ℎ/2 in the middle of the microchannel with inlet velocity of 5 m/s and heat flux of 100 W/cm2 for: (a) hexahedral; and (b) tetrahedral meshes. ... 163

Fig. 6.14: Temperature contours (o_{C) on the 𝑥 − 𝑦 section at 𝐻} 𝑐ℎ/2 at velocity inlet of 5 m/s and heat flux of 100 W/cm2 for: (a) hexhedral; and (b) tetrahedral meshes. ... 163

Fig. 6.15: Schematic of unit cell of rectangular microchannel with boundary condition adopted from the work of Qu & Mudawar (2002b). ... 164

Fig. 6.16: Comparison between the numerical of the present work, Kawano et al. and Qu & Mudawar works for outlet thermal resistance. ... 164

Fig. 6.17: Comparison between the measured and numerical predictions total pressure drop with different 𝑄_𝑖𝑛. ... 165

Fig. 6.18: Comparison between the measured and numerical predictions of total pressure drop with different 𝑅𝑒. ... 166

Fig. 6.19: The total pressure drop for (a) SPSM; (b) DPSM and TPSM models at different water flow rate and at input power of 100 W. ... 167

Fig. 6.20: The total pressure drop in SPSM design at different water flow rate and at input power of 100 W; (a) pressure distribution through 12-serpentine channels; and (b) pressure drop percentage... 168

Fig. 6.21: Pressure drop contours (Pa) of four MCHS at the mid-depth plane of the channel (𝑍 = 𝐻𝑐ℎ/2) : (a) SRM; (b) SPSM ; (c) DPSM; (d) TPSM. ... 169

Fig. 6.22: Distribution of base and fluid bulk temperature along the minichannel axis distance for SRM design for 𝑄_𝑖𝑛 = 0.2 𝑙/𝑚𝑖𝑛 at input power of 100 W. ... 170

Fig. 6.23: Thermocouple temperature readings at different locations inside MCHS versus Reynolds number for input power of 100 W. ... 170

Fig. 6.24: The average Nusselt number versus water flow rate (𝑄_𝑖𝑛) for four MCHS designs at input power of 100 W. ... 171

Fig. 6.25: Temperature contours (oC) of the four MCHSs: (a) SRM; (b) SPSM; (c) DPSM; (d) TPSM. ... 172

(15)

[xiv]

Fig. 6.26: Temperature contours ( C) of the SRM heat sink design. ... 173

Fig. 6.27: Temperature contours (oC) of the SPSM heat sink design. ... 173

Fig. 6.28: Temperature contours (o_{C) of the DPSM heat sink design. ... 174}

Fig. 6.29: Temperature contours (oC) of the TPSM heat sink design. ... 174

Fig. 6.30: (a) Temperature distribution (oC); (b) velocity distribution (m/s) and (c) velocity vectors (m/s) for the SPSM design at 𝑄𝑖𝑛= 0.15 𝑙/𝑚𝑖𝑛 and input power of 100 W. ... 175

Fig. 6.31: Numerical mesh for a SRM design. ... 176

Fig. 6.32: Total thermal resistance versus volumetric flow rate at input power of 100 W. 177 Fig. 6.33: Distribution of Uniform Optimal Latin hypercube DoE points used for metamodel building, 30 points. ... 178

Fig. 6.34: Response surface function using MLSM of total thermal resistance for SPSM heat sink model. ... 179

Fig. 6.35: Response surface function using MLSM of total pressure drop for SPSM heat sink model... 179

Fig. 6.36: Global design trends obtained from the metamodels. ... 180

Fig. 6.37: Pareto front showing the compromises that can be struck in minimising both 𝑅𝑡ℎ and ∆𝑃 together with six representative design points (e.g. P1 ,…, P6) used for the minichannel performance analysis illustrated in Table 6.4. ... 180

Fig. 7.1: 3-D isometric actual and top view of (a) serpentine rectangular MCHS with plate fins (SMPF); (b) serpentine rectangular MCHS with chevron fins (SMCF), all dimensions in mm. ... 183

Fig. 7.2: Schematic of a louvered-fin array, Suga and Aoki (1995). ... 185

Fig. 7.3: Exploded view of serpentine MCHS model with chevron fins. ... 186

Fig. 7.4: Top view of two chevron fins to explain the parameters used for calculating 𝐴_𝑒𝑓𝑓. ... 188

Fig. 7.5: 3-D view and back side of a SMCF design used in simulation to explain the boundary conditions; a) Isometric view; b) Bottom side of the MCHS. ... 190

Fig. 7.6: Grid independent mesh for a SMCF design using grid 3. ... 191

Fig. 7.7: Enlarged view of the computation domain used in the present validation with Lee et al. (2012). ... 192

Fig. 7.8: Results of validation with experimental and numerical study of Lee et al. (2012). ... 193

Fig. 7.9: Total pressure drop versus volumetric flow rate for both serpentine MCHSs proposed at input power of 100 W. ... 193

Fig. 7.10: Total thermal resistance versus volumetric flow rate for both serpentine MCHSs proposed at input power of 100 W. ... 194

Fig. 7.11: Comparison between the experimental total thermal resistance and three components of thermal resistance suggested by Phillips (1987), versus 𝑅𝑒 for: (a) SMPF heat sink; and (b) SMCF heat sink at heat flux of 31 W/cm2 per heater. ... 195

Fig. 7.12: Pressure contours (Pa) for both serpentine MCHSs proposed at 𝑄_𝑖𝑛 of 0.159 l/min and heat flux of 31 W/cm2 per heater; (a) SMCF heat sink; (b) SMPF heat sink. ... 196

(16)

[xv]

Fig. 7.13: Temperature contours ( C) on the 𝑥 − 𝑦 section at 𝐻𝑐ℎ/2 for both serpentine MCHSs proposed at 𝑄_𝑖𝑛 of 0.159 l/min and heat flux of 31 W/cm2 per heater; (a) SMCF heat sink; (b) SMPF heat sink. ... 197 Fig. 7.14: Average channel base temperature versus different Reynolds number for both serpentine MCHSs proposed at heat flux of 31 W/cm2_{per heater. ... 197}

Fig. 7.15: Average Nusselt numbers versus Reynolds number for both serpentine MCHSs proposed at heat flux of 31 W/cm2 per heater. ... 198 Fig. 7.16: : Bar chart and top view of a) amount of the secondary flow diverted from the main minichannel to the secondary microchannel at different 𝜃 with 𝑄_𝑖𝑛 of 0.2 l/min and heat flux of 100 W/cm2; and b) top view to explain the location of the secondary

microchannels (SMC). ... 199 Fig. 7.17: Velocity vectors (m/s) for SMCF models with three different 𝜃 proposed at 𝑄_𝑖𝑛 = 0.2 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2

. ... 200 Fig. 7.18: Velocity vector (m/s) distribution along the streamwise at 𝑄_𝑖𝑛= 0.2 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2_{. ... 201}

Fig. 7.19: Total pressure drop and total thermal resistance at different 𝜃 in a SMCF design with 𝑄𝑖𝑛 = 0.2 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2. ... 201 Fig. 7.20: Total pressure drop and total thermal resistance at different 𝑊𝑠𝑐 in a SMCF design with 𝑄𝑖𝑛 = 0.2 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2. ... 202 Fig. 7.21: Variation in pressure drop through SMCF with different applied heat flux and Reynolds number. ... 203 Fig. 7.22: Maximum surface temperature versus Reynolds number at different heat flux. 204 Fig. 7.23: Variation of 𝐸_𝑁𝑢, 𝐸_∆𝑃 and 𝑃_𝑓 versus 𝜃 with 𝑄_𝑖𝑛 = 0.138 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2_{. ... 205}

Fig. 7.24: Variation of 𝐸𝑁𝑢, 𝐸∆𝑃 and 𝑃𝑓 versus 𝛽 with 𝑄𝑖𝑛 = 0.138 𝑙/𝑚𝑖𝑛 and heat flux of 100 W/cm2. ... 206 Fig. 7.25: Distribution of uniform Optimal Latin Hypercube DoE points used for

metamodel building, 50 points. ... 207 Fig. 7.26: MLS method response surfaces of total thermal resistance for the SMCF heat sink model. ... 208 Fig. 7.27: MLS method response surfaces of total pressure drop for the SMCF heat sink model. ... 209 Fig. 7.28: Global design trends obtained from the metamodels. ... 210 Fig. 7.29: Pareto front showing the compromises that can be struck in minimising both 𝑅_𝑡ℎ and ∆𝑃 together with six representative design points (e.g. P1 ,…, P6) used for the

minichannel performance analysis illustrated in Table 7.5. ... 210 Fig. 7.30: Isometric views of temperature and pressure drop for SMCF heat sink model with 𝑄𝑖𝑛 = 0.16 𝑙/𝑚𝑖𝑛 and constant heat flux of 75 W/cm2; (a) P6 ; and (b) P1 from Table 7.5.

... 212 Fig. 8.1: (a) Layout of a representative GaN HEMT type CREE CGHV1J070D (CREE, 2012); (b) Top view of transistor layout, showing multi-fingered configurations. Source (S), gate (G), and drain (D) metallizations are indicated. All dimensions in µm. ... 215

(17)

[xvi]

(a) Conjugate heat transfer of the MCHS; (b) Isometric view; (c) Bottom side of the MCHS; and (d) the finite element mesh using grid 4 as shown in Fig. 8.3. ... 216 Fig. 8.3: Grid independence test for serpentine and conventional MCHS at 𝑄𝑖𝑛 =

0.16 𝑙/𝑚𝑖𝑛 and total power of 210 W. ... 217 Fig. 8.4: Validation of the current numerical simulation against experimental work of Han et al. (2014) for (a) maximum transistor temperature at different total heating power; (b) temperature distribution along the transistors. ... 219 Fig. 8.5: Comparison for maximum temperature between using a serpentine and a straight MCHS with and without graphene heat spreader versus different 𝑄_𝑖𝑛 and total power of 210 W. ... 220 Fig. 8.6: Comparison for pressure drop between using a serpentine and a straight MCHS with and without graphene heat spreader versus different 𝑄𝑖𝑛 and total power of 210 W. 221 Fig. 8.7: Effect of the 𝑄𝑖𝑛 on maximum heater temperature at different heat spreaders (Graphene, Diamond, SiC, Si) and without spreader, at total power of 210 W and

𝑡𝑠𝑝𝑟𝑒𝑎𝑑𝑒𝑟 = 300 𝜇𝑚. ... 221 Fig. 8.8: Heat flux distribution (W/m2_{) on the top surface of the serpentine MCHS with}

GaN heaters on different heat spreaders: (a) diamond; and (b) graphene, at 210 W, 𝑄_𝑖𝑛 = 0.16 𝑙/𝑚𝑖𝑛 and 𝑡𝑠𝑝𝑟𝑒𝑎𝑑𝑒𝑟 = 300 𝜇𝑚. ... 222 Fig. 8.9: Simulation temperature profile vertically across the structure for heat sink (a) with graphene heat spreader; and (b) with diamond heat spreader; at 𝑄_𝑖𝑛 = 0.16 𝑙/𝑚𝑖𝑛 and 𝑡_{𝑠𝑝𝑟𝑒𝑎𝑑𝑒𝑟} = 300 𝜇𝑚. ... 223 Fig. 8.10: Temperature profile in the longitudinal direction across all gate fingers of last GaN structure at total power of 210 W, 𝑄_𝑖𝑛 = 0.16 𝑙/𝑚𝑖𝑛 and 𝑡_{𝑠𝑝𝑟𝑒𝑎𝑑𝑒𝑟} = 300 𝜇𝑚. ... 224 Fig. 8.11: Effect of the heat spreader thickness on the thermal performance of the structure for four different heat spreaders (Graphene, Diamond, SiC and Si). ... 224 Fig. 9.1: Schematic diagram of the recirculating cooling system... 231

(18)

[xvii]

Nomenclature

𝐴_{𝑏𝑎𝑠𝑒} Base area of minichannel [m2]

𝐴_𝑐ℎ Cross-sectional area of minichannel [m2]

𝐴𝑒𝑓𝑓 Effective heat transfer area per minichannel [m2]

𝐴_𝑓𝑖𝑛 Surface area of fin [m2_]

𝐴_ℎ Bottom heated area of the MCHS [m2]

𝐴_𝑃 Plenum area [m2]

𝐴_{𝑡𝑢𝑏𝑒} Tube area [m2_]

𝐶𝑝_𝑓 Specific heat of fluid [J/kg.K]

𝐷_{𝑡𝑢𝑏𝑒} Tube diameter [m]

𝐷_ℎ Hydraulic diameter [m]

𝑑𝑙 Thickness of thermal layer Ethoxy [m]

𝐅 Body force vector per unit volume [N/m3]

𝑓 Friction factor

𝑓𝑐ℎ Fanning friction factor in minichannel

𝑓𝑎𝑝𝑝 Apparent friction factor

𝐠 Gravity vector [m2/s], 𝑔_𝑥𝑖 + 𝑔_𝑦𝑗 + 𝑔_𝑧𝑘

𝐺 The geometric parameter [∝

2_{+ 1} (∝+1)2]

𝐻_𝑐ℎ Minichannel height [m]

ℎ Convective heat transfer coefficient [W/m2.K]

ℎ_𝑥 Local heat transfer coefficient [W/m2_.K]

𝐼 Current [A]

𝐈 The unit diagonal matrix

𝑘 Turbulent kinetic energy [m2_/s2_]

𝑘_𝑓 Thermal conductivity of fluid [W/m.K]

𝑘𝑙 Thermal conductivity of Ethoxy layer [=2.2 W/m.K]

𝑘_𝑠 Thermal conductivity of copper block [W/m.K]

𝑘_𝑇 Turbulent thermal conductivity of fluid [W/m.K]

𝐿 Heat sink Length [m]

𝐿_𝑐ℎ Minichannel length [m]

𝐿_ℎ𝑦 Hydrodynamic entry length [m]

𝐿_𝑡ℎ Thermal entry length [m]

𝐿_{𝑡𝑢𝑏𝑒} Tube length [m]

(19)

[xviii]

𝑙_𝑠𝑐 Secondary microchannel length [m]

𝑚̇ Mass flow rate [kg/s]

𝑛 Number of minichannel

𝑁_𝑐𝑓 Number of chevron fin

𝑁𝑢 Nusselt number

∆𝑃_𝑐ℎ Minichannel pressure drop [Pa]

∆𝑃_𝑡 Total pressure drop [Pa]

𝑃_𝑐𝑓 Perimeter of chevron fin [m]

𝑝𝑓 Fin pitch (= 𝑙𝑓 + 𝑙𝑠𝑐) [m]

𝑃_𝑖𝑛 Inlet pressure [Pa]

𝑃𝑜𝑢𝑡 Outlet pressure [Pa]

𝑃𝑜 Poiseuille number [=𝑓 ∙ 𝑅𝑒]

𝑃𝑟 Prandtl number

𝑃𝑟_𝑇 Turbulent Prandtl number

𝑃_𝑤 The wetted perimeter [m]

𝑄_𝑖𝑛 Volumetric flow rate [m3_/sec]

𝑞 Heat transfer rate [W]

𝑞𝑖𝑛 Input power to the heater [W]

𝑞_{𝑙𝑜𝑠𝑠} Heat loss [W]

𝑅𝑒 Reynolds number

𝑅_𝑡ℎ Total thermal resistance [K/W]

𝑇_{𝑤,𝑡𝑐𝑖} Minichannel base temperature at thermocouple location [oC]

𝑇𝑓,𝑜𝑢𝑡 Outlet fluid temperature [oC]

𝑇_{𝑓,𝑖𝑛} Inlet fluid temperature [o_C]

𝑇_𝑓,𝑥 Local fluid bulk temperature [oC]

𝑇_{𝑓,𝑎𝑣𝑔} Average fluid bulk temperature [oC]

𝑇_{𝑤,𝑎𝑣𝑔} Average minichannel base temperature [oC]

u Velocity vector [m/s], 𝑢𝑖 + 𝑣𝑗 + 𝑤𝑘

𝑢, 𝑣, 𝑤 Velocities in 𝑥−, 𝑦 − and 𝑧 −directions, respectively [m/s]

𝑈_𝑖𝑛 Inlet water velocity [m/s]

𝑉 Voltage [V]

𝑉_𝑐ℎ Velocity in minichannel [m/s]

𝑉_𝑝 Velocity in plenum [m/s]

(20)

[xix]

Greek symbols

∝ Aspect ratio [= 𝑊_𝑐ℎ/𝐻_𝑐ℎ], 𝑊_𝑐ℎ < 𝐻_𝑐ℎ (∝ < 1)

∝∗ _{Inverse aspect ratio [1 ∝}_{⁄ = 𝐻}

𝑐ℎ/𝑊𝑐ℎ], 𝑊𝑐ℎ < 𝐻𝑐ℎ (∝ > 1)

𝛽 Fin spacing ratio [= 𝑊𝑤/𝑊𝑐ℎ]

𝜉 The excess loss coefficient of bend

𝜂_𝑓 Fin efficiency

𝜌_𝑓 Fluid density [kg/m3]

𝜇𝑓 Dynamic viscosity of fluid [kg/m.s]

𝜇𝑇 Turbulent viscosity of fluid [kg/m.s]

𝜈_𝑓 Kinematic viscosity of fluid [m2_/s]

𝜃 The chevron oblique angle [Degree]

𝜀 Minichannel surface roughness [μm]

𝜏_𝑖𝑗 Viscous stress tensor [Pa]

∇ Del or Nabla (gradient operator),

𝜕 𝜕𝑥𝑖 + 𝜕 𝜕𝑦𝑗 + 𝜕 𝜕𝑧𝑘

Ф The viscous dissipation function

Γ The interface surface between the solid and fluid

𝛿_𝑢 Hydrodynamic boundary layer thickness [m]

𝛿𝑡ℎ Thermal boundary layer thickness [m]

𝜔 Specific dissipation rate [1/sec]

𝜅(𝑥) The Hagenbach factor

𝜅𝑐 Contraction loss coefficients

𝜅_𝑒 Expansion loss coefficients

𝜅₉₀ The bend loss coefficient, (≈ 1.2)

𝑊 Heat sink width [m]

𝑊𝑐ℎ Minichannel width [m]

𝑊_𝑠𝑐 Secondary microchannel width [m]

𝑊_𝑤 Fin width [m]

𝑥, 𝑦, 𝑧 Cartesian coordinates

𝑥+ _{Dimensionless hydrodynamic axial distance [𝑥}+₌ 𝑥

𝑅𝑒∙𝐷ℎ]

𝑥∗ _{Dimensionless thermal axial distance [𝑥}∗₌ 𝑥

(21)

[xx]

Abbreviations

AMD Advanced Micro Devices

CFD Computational Fluid Dynamics

CPUs Central Processing Units

DoEs Design of Experiments

DPSM Double Path Serpentine Minichannel

FDM Finite Difference Method

FEM Finite Element Method

FVM Finite Volume Method

GaN Gallium Nitride

HEMTs High Electron Mobility Transistors

IC Integrated Circuit

ITRS The International Technology Roadmap for Semiconductors

MCHS Minichannel Heat Sink

MLS Moving Least Squares

MMICs Monolithic Microwave Integrated Circuits

RF Radio Frequency

Si Silicon

SiC Silicon Carbide

SMCF Serpentine Minichannel with Chevron Fin

SMPF Serpentine Minichannel with Plate Fin

SPSM Single Path Serpentine Minichannel

SRM Straight Rectangular Minichannel

TPSM Triple Path Serpentine Minichannel

(22)

[1]

Fig. 1.1: shows the increase in the number of transistors every year. The blue line represents Moore’s law applied, starting from the first microprocessor on the graph. The red line

denotes the least square fit of the entire set (Saenen, 2013).

Chapter 1: Introduction

1.1 Thermal Management of Microelectronic Devices

The inexorable miniaturisation of electronic components and increase in electronic packaging density (number of transistors per unit area or volume) is leading to a significant rise in the heat dissipation that must be achieved for operation and to preserve component lifespan and reliability. Hence, it becomes necessary to remove high heat flux from highly compact systems such as high-performance computer chips, laser diodes, gallium nitride (GaN) high electron mobility transistors (HEMTs) and nuclear fusion and fission reactors for ensuring their consistent performance with long life.

Since the invention of the integrated circuit (IC) in the late 1960s, the number of transistors on a chip has increased rapidly (ITRS, 2015). In 1965, Intel co-founder Gordon Moore demonstrates in an empirical relation called Moore's law that the density of components that can be integrated onto a microchip doubles roughly every two years (Moore, 1965). This law has held true since the IC's invention even today as shown in Fig. 1.1.

With growing use of digital systems, the requirements of data storage have grown exponentially and hence, the size and energy consumption of data centres, that house a large part of the digital infrastructure, has greatly increased (Kant, 2009 and Baliga et al., 2011).

(23)

[2]

In 2010, cooling of electronics in data centres account for roughly 33% of 1.31% (238 billion kWh annually) of combined worldwide energy consumption, representing a growth of roughly 11% per year over the last decade (Sharma et al., 2015). For example, in that year the United States (U.S.) data centres consumed about 1.8% of total U.S. electricity consumption, which represents about 82 billion kWh at an annual cost of approximately $6.1 billion (Koomey, 2011; EIA, 2011; and Sverdlik, 2016), and this is expected to triple by 2020 (Shehabi et al., 2016). Assuming an annual increase in new energy generation capacity of 1%, data centres could eventually consume all the available electrical energy in the U.S. by 2030 (Utsler, 2010). Therefore, managing power consumption and energy efficiency of data centres has become essential.

The International Technology Roadmap of Semiconductors (ITRS, 2011) predicted that the power dissipation from a microprocessor chip will exceed 270 W by 2015, and it is expected to increase three times (> 800 W) by 2026. Improving the thermal management in electronic devices is one of the major goals in the development of next generation electronic systems such as central processing units (CPUs), Radio Frequency (RF) Power Amplifier, and Satellite and Radar systems.

Conventional macroscopic air-convection fin-array heat sinks are no longer adequate for these levels of heat generation (Tullius et al., 2011), and in order to dissipate higher heat fluxes (>100 W/cm²), a very high air velocity or a significantly larger-dissipation area is need. For example, Khodabandeh and Palm (2002) claimed that if it is necessary to cool a device component contain a large number of transistors, which dissipates at least 90 W/cm² by using air-cooling heat sink type, a heat sink about 2000 times larger than the area of the device itself is required. Conventional two-phase cooling technologies using heat pipes and vapour chambers have been attractive for microprocessor cooling, because they do not need a pump to rotate the working fluid contained within a closed chamber. However, these capillary-driven devices are not well suited for solving the problem of chip powers exceeding a 100 W, due to limitations in the wick thickness and cross-sectional area of the pipe (Jiang et al., 2002). Even with pulsating and loop heat pipes, the maximum achievable heat flux is still modest as stated by Karayiannis and Mahmoud (2017).

Liquid jet impingement cooling technology, on the other hand, has also attracted much attention from researchers due to its ability to dissipate high heat fluxes. For example, Mudawar et al. (2009) reported 150–200 W/cm2_{using a dielectric fluid HFE-7100 and Zhao}

et al. (2013) reported 674 W/cm2 using water impingement over a porous structure. However, this technology requires high pumping power and there is a danger of surface erosion, due to

(24)

[3]

integration of spray cooling into a closed loop system configuration (Silk et al., 2008 and Karayiannis and Mahmoud, 2017).

For high heat fluxes (> 100 W/cm2_{), single-phase liquid cooling and flow boiling (two-phase}

flow) choices in microfluidic systems can provide the required cooling for the microelectronic devices (Kandlikar et al., 2006). The former, first introduced by Tuckerman and Pease in 1981, have emerged as viable cooling devices for high heat flux electronics due to their light weight, ease of implementation, compactness and higher heat transfer surface area to fluid volume ratio (Yadav et al., 2016). Leonard and Phillips (2005) explained that the use of such heat sinks for cooling of chips could produce savings in energy consumption of over 60%. However, high pressure drop and temperature gradients need to be alleviated. The latter have also been widely studied by researchers due to their ability to dissipate high heat fluxes with much lower pumping power than the former (Ghani et al., 2017a), due to their effective utilization of the latent heat of vaporization of the liquid (Karayiannis and Mahmoud, 2017). However, at higher heat fluxes, microchannel flow boiling suffers from pressure fluctuations and flow instabilities, which can lead to serious problems from significant reductions in heat transfer performance due to, for example, liquid dry-out (Kandlikar et al., 2006 and Balasubramanian et al., 2013).

Fig. 1.2 shows the variation of convective thermal resistances, 𝑅_{𝑐𝑜𝑛𝑣} (K/W), with different coolants and heat transfer mechanisms for a typical heat source area of 10 cm2 and a velocity range of 2–8 m/s. It is shown that when air is used as a coolant the 𝑅_{𝑐𝑜𝑛𝑣} values decreased from 100 K/W to 33 K/W when using natural and forced convection heat transfer mechanisms, respectively, while using fluorocarbon liquids coolant with boiling heat transfer the 𝑅_{𝑐𝑜𝑛𝑣} value decreased significantly to 0.5 K/W.

Fig. 1.2: Typical convective thermal resistances for a 10 cm2_{heat source area (Tummala, 2001).}

Forced Convection Natural Convection Boilin g Air (1–3 atm) Fluorochemical Vapor Silicon Oil Transformer Oil Fluorochemical Liquids Air (1–3 atm) Fluorochemical Vapor Transformer Oil Fluorochemical Liquids Water Water Fluorochemical Liquids 10 cm2_area 0.0 1 0.1 1 10 10 0 100 0 𝑅𝑐𝑜𝑛𝑣 [K/W]

(25)

[4] 1.2 GaN-Based HEMT Technology

Nowadays, most of traditional integrated circuit technologies and power devices are made of silicon (Si) semiconductors, which are not able to operate at temperatures above 250 °C, especially when high operating temperatures are combined with high-power, high-frequency and high-radiation environments (Seal and Mantooth, 2017). As an alternative, Wide Bandgap (WBG) semiconductors, particularly GaN HEMTs and silicon carbide (SiC), have attracted a lot of attention and are considered to be candidates for the next generation high-power electronics such as radio frequency (RF) high-power amplifier implementations and logic applications, as compared to silicon (Si) and gallium arsenide (GaAs)-based devices, due to their superior physical properties, such as high breakdown voltage and capability to handle high current and high power densities (Pengelly et al., 2012). However, the very high-power density in the active region of GaN HEMTs leads to significant degradations in performance as the device temperature increases to more than 300 °C. Thus, effective thermal management of GaN-based electronics is a key to enabling the technology to reach its full potential. Commenting on what was mentioned above, Mishra and Shen (2008) stated that the GaN HEMT technology is becoming the front-runner to be used for power amplifiers, because of its ability to work under robust environments such as higher power levels, high-voltage and high temperatures. Recently, GaN has been used in most power electronic applications, such as: radar and space applications, missiles, satellites, automotive industry, defence and military communications, high frequency Monolithic Microwave Integrated Circuits (MMICs), high power amplifiers for wireless base stations, high voltage electronics for power transmission lines, cell phone infrastructure (base stations), as well as automotive and other general power conversion devices (Flack et al., 2016). This is attributed to a unique combination of GaN material properties, including wide bandgap (3.4 eV), high saturation electron drift velocity (2.5×107_{cm/s) and large electric breakdown field strengths (~3.3×10}6_{V/cm). These}

properties are explained in Table 1.1 (Kaminski and Hilt, 2014), which compares the benefits of GaN to those of other major and relevant semiconductors.

Table 1.1: Comparison between GaN and other Semiconductor properties (Kaminski and Hilt, 2014).

Property Si SiC GaAs GaN Diamond

Bandgap, Eg (eV) 1.12 3.26 1.43 3.39 5.47

Electric Breakdown Field, Ecrit

(MV/cm) 0.23 2.2 0.4 3.3 5.6

Electron Mobility, 𝜇𝑛 (cm2/V.s) 1400 950 8500 1500 1800

Permittivity, 𝜀𝑟 11.8 9.7 13.1 9.0 5.7

Saturated Electron Drift Velocity, 𝑣𝑠𝑎𝑡 (×107 cm/sec)

1 2 1 2.5 2.7

Thermal Conductivity, k (W/cm.K) 1.5 3.8 0.46 1.3 20

(26)

[5]

As can be seen, GaN has superior physical properties compared with other semiconductors, especially the conventionally used Si, for power devices operating under power, high-temperature and high-frequency conditions. For this reason, GaN HEMTs have proven to be excellent candidate for the radio frequency (RF) power amplifier applications (Nuttinck et al., 2002).

Although GaN has so many advantages that make it a promising technology, significant technological development is still required to improve the reliability and thermal management of devices constructed with this material. Self-heating is one of the critical factors that adversely affect device performance and reliability. Chattopadhyay and Tokekar (2008) claimed that GaN HEMT devices offer high power density, high break-down voltage and therefore generate high chip temperatures that can reach several hundred degrees above ambient base temperature due to self-heating, and their study has demonstrated that increasing the chip temperature not only reduces the performance of the devices but also accelerates their degradation. Severe self-heating effects may damage the gate electrode and can burn metal wires connecting the chip to the package, and hence cause device failures and reliability problems.

To avoid the malfunctions of electronics and to ensure the reliability of the electronic systems, with the rapid increase of heat dissipation from various electronic components such as CPU and graphic card in computers and radio-frequency (RF) power amplifiers, substantial research work has been carried out to explore more effective cooling techniques to keep up with the development pace of new electronic equipment and large electronic systems.

1.3 Liquid Cold Plate Microchannel Heat Sink

Nowadays, the thermal designer faces many challenges to overcome the factors of contracting system size, the requirement to insert more components within a limited space and to reduce the system acoustic noise generated from the heat sinks fans. The thermal solution is required to dissipate the maximum power consumption of the electronic equipment and ensure it remains below its maximum operating temperature (junction temperature). To overcome this dilemma, there are several liquid cooling methods, among these direct and indirect contact liquid cooling techniques. One example of direct contact liquid cooling is that investigated by Almaneea (2014) through immersing servers into dielectric liquid. The indirect contact liquid cooling method, which is the subject of this study, uses a liquid cold plate microchannel heat sink in an open-loop single-phase system, which is a promising solution to thermal challenges (Fan et al., 2014).

(27)

[6]

Liquid cooling like pure water, dielectric (Hydrofluoroethers, HFE) and nanofluids (nanofluid is a mixture between water and organic material such as Al2O3, SiO2, TiO2 and

CuO) have much higher thermal conductivities and specific heat capacities than air, and thus much higher heat transfer coefficients associated with them. Therefore, by using liquids for cooling of electronic components is more effective than gas or air cooling (Salman et al., 2014).

Microchannel heat sinks can be used as an effective heat dissipation device, which has proven to be a very efficient method to remove high heat loads generated from a chip (Chein and Chuang, 2007). The novel idea of dispelling heat through the use of a silicon based MCHS was first introduced by Tuckerman and Pease in the early 1980s (Tuckerman and Pease, 1981) as shown in Fig. 1.3, which consists of an array of straight rectangular microchannels etched in a 1 cm2_{silicon wafer. They pointed out that decreasing liquid cooling channel dimensions}

to the micron scale (in the range of 10 μm to 103 μm) will lead to an increase in the heat transfer rate. The hydraulic diameter, 𝐷_ℎ (m), of the microchannel is inversely proportional to the heat transfer coefficient within a narrowed channel of laminar flow, where the pressure drop is proportional to 𝐷_ℎ−4_{(Bergman et al., 2017). Hence, a higher heat transfer coefficient} can be achieved by decreasing the hydraulic diameter of the channels at the expense of increase in pressure drop.

For channels of rectangular cross-section, the hydraulic diameter (𝐷_ℎ) is defined as:

𝐷ℎ= 4 ∙ 𝐴𝑐ℎ 𝑃𝑤 =2(𝑊𝑐ℎ∙ 𝐻𝑐ℎ) 𝑊𝑐ℎ+ 𝐻𝑐ℎ (1.1) 𝑊

𝑞

Cover plate 𝑦 𝑧 𝑥 𝐻𝑐ℎ 𝐻

𝑞

𝐻𝑏 Silicon 𝑊_𝑐ℎ+ 𝑊_𝑤 𝑊_𝑤 2 𝑊_𝑐ℎ Cover plate 𝐿

(28)

[7]

where 𝐴_𝑐ℎ is the channel cross sectional area (m2) and 𝑃_𝑤 is the channel wetted perimeter (m), while 𝑊_𝑐ℎ and 𝐻_𝑐ℎ represent the width and depth of the channel (m), respectively. After this pioneering work, several studies have investigated the fluid flow and heat transfer characteristics of the microchannel heat sink, with a comprehensive review being found in Ghani et al. (2017a).

1.4 Classification of the Microchannels

Classification of the channel size is a controversial issue, and several authors have addressed this problem. Mehendale et al. (2000) had classified the channels according to their dimension, "𝐷ℎ", being the hydraulic diameter of tube or smallest dimension for other cross-sections. Kandlikar and Grande (2003) had distinguished the channels based on fluid flow which they recommend for both gas and liquid (single-phase flow) as well as two-phase flow applications. The nomenclatures adopted by both these authors are presented in Table 1.2.

Obot (2003) classified channels of hydraulic diameter under 1 mm (𝐷_ℎ ≤ 1 mm) as microchannels, which was also adopted by many other researchers such as Bahrami et al. (2006) and Bayraktar and Pidugu (2006).

As a consequence, distinguishing between macrochannel and microchannel flows cannot be based on the channel size alone, for this reason Kew and Cornwell (1997) recommend using a confinement number (𝐶𝑜) as a transition criterion between macro and microscale flow boiling (two phase flow) as given by Eq. (1.2):

𝐶𝑜 = ( 𝜎 𝑔(𝜌𝑙− 𝜌𝑣)𝐷ℎ2 ) 1 2 (1.2)

where 𝜎 represents the surface tension (N/m), while 𝑔 is the acceleration due to gravity (m/s2

). 𝜌𝑙 and 𝜌𝑣 are densities for liquid and vapour case (kg/m3), respectively. As per their proposed criteria, they observed that the transition happened at 𝐶𝑜 = 0.5, with 𝐶𝑜 < 0.5 being macro-scale flow and 𝐶𝑜 > 0.5 as microchannel flow.

Mehendale et al. (2000) Kandlikar and Grande (2003)

Conventional Channels 𝐷ℎ > 6 mm Conventional Channels 𝐷ℎ > 3 mm Compact Passages 1 mm < 𝐷ℎ ≤ 6 mm Minichannels 200 μm < 𝐷ℎ ≤ 3 mm Meso-channels 100 μm < 𝐷ℎ ≤ 1 mm Microchannels 10 μm < 𝐷ℎ ≤ 200 μm Micro-channels 1 μm < 𝐷ℎ ≤ 100 μm Transitional Channels 0.1 μm < 𝐷ℎ ≤ 10 μm

Transitional Microchannels 1 μm < 𝐷ℎ ≤ 10 μm Transitional Nanochannels 0.1 μm < 𝐷ℎ ≤ 1 μm Molecular Nanochannels 𝐷ℎ ≤ 0.1 μm

(29)

[8]

In the present study, the criteria proposed by Kandlikar and Grande (2003) has been adopted here. Since the channel hydraulic diameter of all the heat sinks fabricated were between the 1.334 mm and 2 mm, minichannels are used for the current work.

1.5 Thermal Resistance Network in Heat Sinks

Generally, the heat generated by the chip mounted on the heat sink base is firstly transferred to the microchannels by conduction and then to the coolant by convection (Fan et al., 2014). A convenient way of presenting the cooling capability of a heat sink for electronics cooling is the total thermal resistance, 𝑅𝑡ℎ (K/W), which is defined as the ratio of the temperature difference across the module from the electronic die to the coolant fluid per watt of energy transferred (Mochizuki et al., 2011), as defined in Eq. (1.3):

𝑅𝑡ℎ=

𝑇𝑚𝑎𝑥− 𝑇𝑓,𝑖𝑛

𝑞 (1.3)

where,

𝑇𝑚𝑎𝑥 = Maximum surface temperature in heat sink [K]. 𝑇𝑓,_𝑖𝑛 = Fluid inlet temperature [K].

𝑞 = Heat dissipation [W]

A low thermal resistance is desirable in order to minimize the temperature rise of the electronic devices per watt of heat generated. The thermal resistance of the system is the sum of each individual layer's thermal resistance, which depends on the thermal conductivity and thickness of the material. A traditionally used cooling technique with forced water convection through the heat sink is schematically presented in Fig. 1.4. In this figure, a single microchannel with symmetry has been chosen to illustrate the thermal resistance network. Shao et al. (2009) carried out a study on the MCHS and considered three main thermal resistances given in Fig. 1.4. The conduction thermal resistances of the bottom base and the fin are represented as 𝑅_{𝑐𝑜𝑛𝑑,}_{𝑏𝑎𝑠𝑒} and 𝑅_{𝑐𝑜𝑛𝑑,}_𝑓𝑖𝑛, respectively. While 𝑅_{𝑐𝑜𝑛𝑣,}_{𝑏𝑎𝑠𝑒} and 𝑅_{𝑐𝑜𝑛𝑣,}_𝑓𝑖𝑛 are respectively the thermal resistances due to convection from the heat sink base and the walls (fins) to the fluid. The last thermal resistance is due to bulk temperature rise in the fluid, 𝑅_{𝑏𝑢𝑙𝑘} (K/W). Hence, the Eq. (1.3) can be modified to written as:

𝑅𝑡ℎ=

𝑇𝑚𝑎𝑥− 𝑇𝑓,_𝑖𝑛

𝑞 = 𝑅𝑐𝑜𝑛𝑑+ 𝑅𝑐𝑜𝑛𝑣+ 𝑅𝑏𝑢𝑙𝑘 (1.4)

where 𝑅_{𝑐𝑜𝑛𝑑} is the equivalent conductive thermal resistances (K/W), and is equal to the network of (𝑅_{𝑐𝑜𝑛𝑑,}_{𝑏𝑎𝑠𝑒}+ 𝑅_{𝑐𝑜𝑛𝑑,}_𝑓𝑖𝑛), while 𝑅_{𝑐𝑜𝑛𝑣,}_{𝑏𝑎𝑠𝑒} + 𝑅_{𝑐𝑜𝑛𝑣,}_𝑓𝑖𝑛 represent the equivalent

(30)

[9]

convective thermal resistances, 𝑅_{𝑐𝑜𝑛𝑣} (K/W). Each thermal resistance listed here will be explained in detail in the subsection 3.8.3.

1.6 Study Objective and Aims

Thermal management is one of the most critical areas for the electronic product development, and a large proportion of field failures can be attributed to overheating, which in turn is caused by inappropriate thermal design (Sahin et al., 2005). The thermal problem has a significant impact on the cost, overall design, reliability and performance of the next generation of microelectronic devices (ITRS, 2015). The conventional straight rectangular microchannel heat sink has a main problem which is the continuous increase of surface temperature along the flow direction (Li et al., 2004).

Therefore, the main objective of this study is to identify, design, fabricate and test a suitable cooling technique capable to dissipate high heat flux (> 100 W/cm2) from a chip. There is an urgent need to find an optimal thermal design of a microchannel heat sink cooled by a low flow rate coolant (water). To overcome this limitation, three different configurations of serpentine minichannel heat sink (MCHS) devices have been designed, fabricated and tested as an option to address the problem of thermal management, which they termed single (SPSMs), double (DPSMs) and triple path multi-serpentine rectangular minichannels (TPSMs) as shown in Fig. 1.5. The performance of these types of the minichannel heat sinks have been compared experimentally and numerically with an array of straight rectangular

Fig. 1.4: Single microchannel with thermal resistance networks.

𝑅𝑐𝑜𝑛𝑣,𝑏𝑎𝑠𝑒 𝑅𝑐𝑜𝑛𝑑,𝑓𝑖𝑛 𝑅𝑐𝑜𝑛𝑑,𝑏𝑎𝑠𝑒 𝑅_{𝑐𝑜𝑛𝑣,𝑓𝑖𝑛} 𝑥 𝑦 𝑧 𝑞 Symmetry 𝑇𝑓,𝑜𝑢𝑡 𝑇𝑏𝑎𝑠𝑒 𝑅𝑐𝑜𝑛𝑑,𝑓𝑖𝑛 𝑅_{𝑐𝑜𝑛𝑣,𝑓𝑖𝑛} 𝑅𝑐𝑜𝑛𝑑,𝑏𝑎𝑠𝑒 𝑅𝑐𝑜𝑛𝑣,𝑏𝑎𝑠𝑒 𝑅𝑏𝑢𝑙𝑘 𝑇𝑓,𝑜𝑢𝑡 𝑇𝑏𝑎𝑠𝑒 𝑊𝑤 2 𝑊𝑐ℎ 2

⟿

(31)

[10]

minichannels (SRMs) in terms of pressure drop (∆𝑃), average Nusselt number (𝑁𝑢_𝑎𝑣𝑔) and total thermal resistance (𝑅_𝑡ℎ).

The key idea behind choosing these types of the MCHSs (Serpentine) in cooling of a chip, that is periodically interrupt hydrodynamic and thermal boundary layers in the corner (bends). In other words, periodic restart in both hydrodynamic and thermal boundary layers hinder the development of the thermal boundary layer on smooth surfaces responsible for limiting the heat transfer rates in MCHS designs. These improvements in heat transfer are, however, achieved at the expense of significantly large increase in ∆𝑃. Additionally, the serpentine MCHS designs have large ratio of heat transfer surface to fluid flow volume compared with the conventional heat sink due to the bends, which leads to absorb as much of a heat from a chip (Lee et al., 2015).

Thermal management in gallium nitride on Silicon carbide (GaN-on-SiC) power amplifier based microelectronic devices is studied numerically. Three-dimensional numerical simulation of the conjugate heat transfer has been carried out within COMSOL Multiphysics. To enhance the hotspot cooling capability of the water-cooled copper SPSM heat sink, four different materials of heat spreader which of Si, SiC, diamond and graphene mounted between the GaN die and the heat sink base are used. The transistor selected for numerical

(a) _(b)

(c) (d)

Fig. 1.5: 3-D schematic view of MCHS suggested; a) SRM; b) SPSM; c) DPSM, d) TPSM heat sink.

(32)

[11]

(a)

(b)

(c)

simulation study was the CREE CGHV1J070D GaN HEMT die, as shown in Fig. 1.6, which is attached directly on the heat spreader.

Three small heaters of size 4.8 × 0.8 mm2 were used to generate a total power of 210 W (power on each GaN is 70 W), resulting in the heat flux on each GaN transistor is 1.823 kW/cm2_{. The effect of the heat spreader thickness on the maximum chip temperature also has}

been investigated.

To enhance the convective heat transfer and achieve a more homogeneous temperature distribution together with reduction in pressure drop penalty, flow obstructions and secondary flows in microchannels can be produced by adding secondary microchannels between the main flow microchannels (Steinke and Kandlikar, 2006a). The present study explores a new design of heat sink where chevrons fins within multiple serpentine minichannels are used to control the hydrodynamic and thermal boundary layers. According to the author's knowledge, no researches have been published on secondary flow using chevron fins with a serpentine MCHS, and this has motivated the present study to develop a SPSM heat sink with chevron fins that could enhance the heat transfer together with significant reduction in pressure drop. Hence, another experimental and numerical work have been carried out to explore the ability of enhancing the heat transfer in SPSM with reduction pressure drop, see Fig. 1.7.

Fig. 1.6: (a) CGHV1J070D GaN-on-SiC HEMT bare die which is inside (b); (b) High voltage GaN HEMT on SiC substrate packaging which is inside (c); (c) Radio

Serpentine minichannel liquid-cooled heat sinks for electronics cooling applications